Bismark CpA dinucleotides

[biocomputer]

I would like to specifically look at CpA dinucleotide methylation (ie. make a bedgraph) like I can with CpG's using Bismark. What is the best way to do this? I think I could run bedGraph2cytosine on the CHH bedgraph then use the 2nd last column of the bedGraph2cytosine output (C-context) to pull out all lines that contain "CA_".

Also just to clarify something something, a CHH file should contain everything in the CpG file + CHG file + other C contexts, or does CHH contain all C contexts excluding CpG and CHG?

[dpryan]

Regarding the CHH file, it will not contain bases in either the CpG or the CHG file, since H means "everything except G".

For getting CpA context methylation, you'd presumably need to go through both the CHH and CHG files (note that you might look at them separately first to ensure that the CHG context doesn't itself have an effect that you need to account for!). Personally I'd just modify PileOMeth to get CpA metrics, but to each their own

[fkrueger]

We have actually done exactly that in the past, so I can just attach a script to do the job. In short you can just run bismark2bedgraph (maybe only using non-CG files from the methylation extraction (since CHG files may also have CA-calls)) using the following command:

bismark2bedGraph --CX -o non-CG.bedGraph CH*

and then proceed further to coverage2cytosine:

coverage2cytosine --genome /your/genome/ -o non-CG.CX_report.txt non-CG.cov.gz

Then you can run the attached script to filter out all CA calls and write out a new CA.cov file with the following command:

filter_out_CA_calls.pl *CX_report.txt

All the best, Felix